Measuring Growth Spillovers

ABSTRACT We propose a multilevel econometric model with time‐varying spillover parameters that disentangle within‐country from between‐country growth spillovers. Parameter estimation is carried out by the method of maximum likelihood. The finite‐sample properties of the resulting estimates are validated through a Monte Carlo study. We illustrate the model properties in an empirical application for six Latin American countries. The results show that our multilevel framework outperforms a benchmark model without the multilevel data structure. Specifically, the benchmark model tends to underestimate the spillover parameter compared to our multilevel approach. Also, the between‐country spillover is subject to cyclical dynamics, varying from negative to positive values, and suggests that growth in partner countries can either accelerate or decelerate their domestic growth. In contrast, all within‐country spillovers are positive and can be regarded as growth multipliers that reinforce internal economic dynamics within each country.